Dyna — Dyna Reinforcement Learning Framework

Fact

A reinforcement learning framework using both real experience and model-made practice.

In Plain Words

Dyna is like driver's ed with a simulator. The student drives one real lap, then does ten pretend laps without hitting real cones.

It helps an agent learn with fewer real-world mistakes. You see it in MBRL and robot control.

Related Concepts

• Model-Based Reinforcement Learning
• Temporal-Difference Learning
• Q-Learning

MBRL
Dyna is an early MBRL framework for planning with a learned model.

TD Learning
Dyna uses TD Learning on both real and simulated experience.

Q-Learning
Dyna-Q adds planning steps to Q-Learning.